There are a couple things I can address in your question. First, the metadata file is not the output of another analysis. This is extra data you should be collecting during your experiment. I usually build mine in excel, but they need to be saved as a .txt file to use in QIIME.
Categorical data is a variable that groups the data, but it isn’t something you can count. For example, in the metadata file Sai attached, there is a column called Body-site and a column called Day. The way it is expressed, day is numeric because you can count, and body site is categorical.
Depending on your research question, you may or may not have categorical data. If you only have numeric metadata, you could consider alpha-correlation instead of alpha-group significance.
Let me know if that doesn’t answer your question